Improved Hilbert transform to process the ultrasonic TOFD signal

The difficulty in the TOFD is to identify the arrival time of every echo in the overlap waves, an improved Hilbert transform is proposed in this paper. Firstly, calculating the energy of the noise in the original signal, then processing the original signal and the pure noise that have the same energy as original signal with the EMD method. Comparing every IMF´s energy of the two signal, we can decide which IMF to reserve and which to eliminate, filtering partial noisy of the original signal. After that, the threshold denoising is made to the selective IMFs to minimize the noise´s impact. Finally, we can perform Hilbert transform. The simulation testified the effectiveness and accuracy of this technique and it can identify the TOFD signal with lower signal-to-noise ratio than perform Hilbert transform directly. In the experiment, the minimum time interval which can identify with this technique is 0.2us, the average error in identify some kinds of typical defects is 0.13mm.